Supplementary Figures – Radiomics of Dual energy CT for predicting progression-free survival in patients with Early Glottic Cancer

Supplementary Figure 1. Radiomics of Dual energy CT for predicting progression-free survival in patients with Early Glottic Cancer

Enrollment Flowchart of patients with EGC in this study. EGC: Early glottic cancer. CT, computed tomography .

Supplementary Figure 2. Radiomics of Dual energy CT for predicting progression-free survival in patients with Early Glottic Cancer

The tumor segmentation of a patient larger than 1cm. Yellow region, the segmented tumors in VP images; green region, the segmented in iodine maps.

Supplementary Figure 3. Radiomics of Dual energy CT for predicting progression-free survival in patients with Early Glottic Cancer

(A) Coefficient profile plot, shrinking the coefficients of radiomic features to 0 to select key features. (B) The tuning parameter () was obtained by the least absolute shrinkage and selection operator (LASSO) model with 10-fold cross-validation through standard error of the minimum criteria. Finally 5 significant features were selected to constructed radiomic risk model (original glcm Cluster Shade, original glcm MCC, wavelet glcm Cluster Shade, HHH glszm Large Area Low Gray Level Emphasis, HHH ngtdm Busyness).

Supplementary Figure 4. Radiomics of Dual energy CT for predicting progression-free survival in patients with Early Glottic Cancer

(A-B) Heatmap showing the 5 eligible radiomic features and distributions of risk score in the train cohort, the blue dot represents patients who are predicted with lower risk and the red dot represents patients with higher risk. (C-D) Heatmap showing the 5 eligible radiomic features and distributions of risk score in the validation cohort.

Supplementary Figure 5. Radiomics of Dual energy CT for predicting progression-free survival in patients with Early Glottic Cancer

Difference between the risk score and T stage in validation set, T2 stage significantly higher than patients with T1 stage(p=0.03). No significant statistics difference of RRS was found between Gender, Age and Therapy groups.